Diagnosis by ocular fundus imaging plays a key role in the monitoring, detection, and diagnosis of several diseases including eye-related, hypertensive, and cardiovascular diseases as well as the detection of brain microvascular and neuronal pathology. The spectral detection of the fundus in the near-infrared (NIR) region can be used to analyze the composition and content of the fundus substance. The aim of this study was to develop a multichannel bandpass filter, which can be assembled on a CMOS image sensor installed in a fundus camera, to realize spectral imaging. We designed the filter in the NIR region according to interference filter theory and fabricated it by sputtering, lithography, and etching. The images acquired by the image sensor with the fabricated bandpass filter confirmed that the pixel values were comparable to the values calculated from spectrometer measurements. In an experimental validation, the number of measurement wavelengths was increased by applying the incident angle dependence of the interference filter. Each fabricated filter had a relatively wide transmittance band for spectral detection, broadening the spectrum curve. Moreover, a high-resolution spectrum was reconstructed using the developed algorithm. The spectrum reconstructed by the proposed filter had a 15% average relative error over the entire wavelength range of interest.